idnits 2.17.1 draft-ietf-v6ops-nd-cache-init-00.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (October 27, 2019) is 1643 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- -- Obsolete informational reference (is this intentional?): RFC 4941 (Obsoleted by RFC 8981) Summary: 0 errors (**), 0 flaws (~~), 1 warning (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 v6ops J. Linkova 3 Internet-Draft Google 4 Intended status: Informational October 27, 2019 5 Expires: April 29, 2020 7 Neighbor Cache Entries on First-Hop Routers: Operational Considerations 8 draft-ietf-v6ops-nd-cache-init-00 10 Abstract 12 Neighbor Discovery (RFC4861) is used by IPv6 nodes to determine the 13 link-layer addresses of neighboring nodes as well as to discover and 14 maintain reachability information. This document discusses how the 15 neighbor discovery state machine on a first-hop router is causing 16 user-visible connectivity issues when a new (not being seen on the 17 network before) IPv6 address is being used. 19 Status of This Memo 21 This Internet-Draft is submitted in full conformance with the 22 provisions of BCP 78 and BCP 79. 24 Internet-Drafts are working documents of the Internet Engineering 25 Task Force (IETF). Note that other groups may also distribute 26 working documents as Internet-Drafts. The list of current Internet- 27 Drafts is at https://datatracker.ietf.org/drafts/current/. 29 Internet-Drafts are draft documents valid for a maximum of six months 30 and may be updated, replaced, or obsoleted by other documents at any 31 time. It is inappropriate to use Internet-Drafts as reference 32 material or to cite them other than as "work in progress." 34 This Internet-Draft will expire on April 29, 2020. 36 Copyright Notice 38 Copyright (c) 2019 IETF Trust and the persons identified as the 39 document authors. All rights reserved. 41 This document is subject to BCP 78 and the IETF Trust's Legal 42 Provisions Relating to IETF Documents 43 (https://trustee.ietf.org/license-info) in effect on the date of 44 publication of this document. Please review these documents 45 carefully, as they describe your rights and restrictions with respect 46 to this document. Code Components extracted from this document must 47 include Simplified BSD License text as described in Section 4.e of 48 the Trust Legal Provisions and are provided without warranty as 49 described in the Simplified BSD License. 51 Table of Contents 53 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 54 1.1. Requirements Language . . . . . . . . . . . . . . . . . . 4 55 1.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 4 56 2. Potential Solutions . . . . . . . . . . . . . . . . . . . . . 5 57 2.1. Do Nothing . . . . . . . . . . . . . . . . . . . . . . . 5 58 2.1.1. Pros . . . . . . . . . . . . . . . . . . . . . . . . 5 59 2.1.2. Cons . . . . . . . . . . . . . . . . . . . . . . . . 5 60 2.2. Change to the Registration-Based Neighbor Discovery . . . 6 61 2.3. Hosts Explicitly Advertizing Their GUAs Using Existing ND 62 Mechanisms . . . . . . . . . . . . . . . . . . . . . . . 6 63 2.3.1. Host Sending Unsolicited NA . . . . . . . . . . . . . 6 64 2.3.1.1. Pros . . . . . . . . . . . . . . . . . . . . . . 7 65 2.3.1.2. Cons . . . . . . . . . . . . . . . . . . . . . . 7 66 2.3.2. Host Sending NS to the Router Address from Its GUA . 7 67 2.3.2.1. Pros . . . . . . . . . . . . . . . . . . . . . . 8 68 2.3.2.2. Cons . . . . . . . . . . . . . . . . . . . . . . 8 69 2.3.3. Host Sending Router Solicitation from its GUA . . . . 8 70 2.3.3.1. Pros . . . . . . . . . . . . . . . . . . . . . . 8 71 2.3.3.2. Cons . . . . . . . . . . . . . . . . . . . . . . 8 72 2.4. Routers Populating Their Caches by Gleaning From Neighbor 73 Discovery Packets . . . . . . . . . . . . . . . . . . . . 9 74 2.4.1. Pros . . . . . . . . . . . . . . . . . . . . . . . . 9 75 2.4.2. Cons . . . . . . . . . . . . . . . . . . . . . . . . 9 76 2.5. Initiating Hosts2Routers Communication . . . . . . . . . 9 77 2.5.1. Pros . . . . . . . . . . . . . . . . . . . . . . . . 10 78 2.5.2. Cons . . . . . . . . . . . . . . . . . . . . . . . . 10 79 2.6. Tweaking Probing Algorithms . . . . . . . . . . . . . . . 10 80 2.7. Routers Buffering More Packets . . . . . . . . . . . . . 10 81 2.7.1. Pros . . . . . . . . . . . . . . . . . . . . . . . . 10 82 2.7.2. Cons . . . . . . . . . . . . . . . . . . . . . . . . 11 83 3. Recommendations . . . . . . . . . . . . . . . . . . . . . . . 11 84 4. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 11 85 5. Security Considerations . . . . . . . . . . . . . . . . . . . 11 86 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 11 87 7. References . . . . . . . . . . . . . . . . . . . . . . . . . 12 88 7.1. Normative References . . . . . . . . . . . . . . . . . . 12 89 7.2. Informative References . . . . . . . . . . . . . . . . . 13 90 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 13 92 1. Introduction 94 The section 7.2.5 of [RFC4861] states: "When a valid Neighbor 95 Advertisement is received (either solicited or unsolicited), the 96 Neighbor Cache is searched for the target's entry. If no entry 97 exists, the advertisement SHOULD be silently discarded. There is no 98 need to create an entry if none exists, since the recipient has 99 apparently not initiated any communication with the target." 101 This approach is perfectly suitable for host2host communications 102 which are in most cases bi-directional and it could be expected that 103 if a host A has an ND cache entry for the host B IPv6 address, the 104 host B also has the corresponding ND entry for the host A address in 105 its cache. However when a host communicates to off-link destinations 106 via its first-hop router that logic does not apply. Here is the most 107 typical scenario when the problem may arise: 109 1. When a host joins the network it receives an RA packet from the 110 first-hop router (either a periodic unsolicited RA or a response 111 to an RS sent by the host). The RA contains information the host 112 needs to perform SLAAC and to configure its network stack. Among 113 other things the host populates its ND cache with the router 114 link-local address and potentially link-layer address (if 115 included in the RA Source Link-Layer Address option). 117 2. The host starts opening connections to off-link destinations. 118 Very common use case is a mobile device sending probes to detect 119 the Internet connectivity and/or the captive portals presence on 120 the network. To speed up that process many implementations are 121 using the Optimistic Duplicate Address Detection ([RFC4429]) 122 which allows them to send probes from their GUA before the DAD 123 process is completed. Imprortant point here is that at that 124 moment the device ND cache contains all information required to 125 send those probes (such as the default gateway LLA and the link- 126 layer address). The router ND cache, however, might contain an 127 entry for the device link-local address (if the device has been 128 performing the ND process for the roiter LLA) but there are no 129 entries for the device GUA. 131 3. Response packets for the probes (or any other traffic sent by the 132 host) are received by the first-hop router. As the router does 133 not have any ND cache entry for the host GUA, the router starts 134 the neighbor discovery process by creating an INCOMPLETE cache 135 entry and then sending an NS to the Solicited Node Multicast 136 Address. Apparently most of the router implementations buffer 137 only one data packet while performing the ND process for its 138 destination. Therefore all packets for the host GUA, except for 139 the very first one are dropped until the address resolution 140 process is completed. 142 4. As many implementations send multiple probes in parallel it's 143 very likely that all probes ex. the first one would be considered 144 failed. If the host implements an exponential backoff for 145 probing it leads to user-noticeable delay in detecting network 146 connectivity/reporting the network as usable. 148 The above-mentioned scenario illustrates the problem happening when 149 the device connects to the network for the first time/after a long 150 timeout. However the same sequence of events happen when the host 151 starts using the new (previously unseen by the router or ) GUA (e.g. 152 a new privacy address [RFC4941]) or if the router Neighbor Cache has 153 been flushed. 155 While in dual-stack networks this problem might hidden by Happy 156 Eyeballs ([RFC8305]) it manifests itself quite clearly in IPv6-only 157 networks, especially wireless ones, leading to poor user experience 158 and contributing to negative perception of IPv6-only solutions as 159 unstable and non-deployable. 161 1.1. Requirements Language 163 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 164 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 165 "OPTIONAL" in this document are to be interpreted as described in BCP 166 14 [RFC2119] [RFC8174] when, and only when, they appear in all 167 capitals, as shown here. 169 1.2. Terminology 171 ND: Neighbor Discovery, [RFC4861]. 173 SLAAC: IPv6 Stateless Address Autoconfiguration, [RFC4862]. 175 NS: Neighbor Solicitation, [RFC4861]. 177 NA: Neighbor Advertisement, [RFC4861]. 179 RS: Router Solicitation, [RFC4861]. 181 RA: Router Advertisement, [RFC4861]. 183 SLLA: Source link-layer Address, an option in the ND packets 184 containing the link-layer address of the sender of the packet 185 ([RFC4861]). 187 TLLA: Target link-layer Address, an option in the ND packets 188 containing the link-layer address of the target ([RFC4861]). 190 GUA: Global Unicast Address ([RFC4291]). 192 DAD: Duplicate Address Detection, [RFC4862]. 194 Optimistic DAD: a modification of DAD, [RFC4429]. 196 2. Potential Solutions 198 The problem could be addressed from different angles. Possible 199 approaches are: 201 o Just do nothing. 203 o Migrate from the "reactive" Neighbor Discovery ([RFC4861]) to the 204 registration-based mechanisms ([RFC8505]). 206 o The host explicitly advertizes its GUAs using Neighbor Discovery 207 mechanisms. 209 o The router creates new entries in its Neighbor Cache by gleaning 210 from Neighbor Discovery DAD messages. 212 o The host initiates bidirectional communication to the router using 213 the host GUA. 215 o Making the probing logic on hosts more robust. 217 o Increasing the buffer size on routers. 219 The following sections discuss those approaches in more detail. 221 2.1. Do Nothing 223 One of the possible approaches might be to declare that everything is 224 working as intended. 226 2.1.1. Pros 228 o No work required. 230 2.1.2. Cons 232 o Unhappy users. 234 o Many support tickets. 236 o More resistance to deploy IPv6 and IPv6-Only networks. 238 2.2. Change to the Registration-Based Neighbor Discovery 240 The most radical approach would be to move away from the reactive ND 241 as defined in [RFC4861] and expand the registration-based ND 242 ([RFC6775], [RFC8505]) used in Low-Power Wireless Personal Area 243 Networks (6LoWPANs) to the rest of IPv6 deployments. 245 This option required some investigation and discussions and seems to 246 be an overkill for the problem described in this document.. 248 2.3. Hosts Explicitly Advertizing Their GUAs Using Existing ND 249 Mechanisms 251 The Neighbor Discovery is designed to allow IPv6 nodes to discover 252 neighboring nodes reachability and learn IPv6 to link-layer addresses 253 mapping. Therefore ND seems to be the most appropriate tool to 254 inform the first-hop routers about addresses the host is going to 255 use. The following sections discuss potential apptoaches in more 256 detail. 258 2.3.1. Host Sending Unsolicited NA 260 Section 4.4 of [RFC4861] says: 262 "A node sends Neighbor Advertisements in response to Neighbor 263 Solicitations and sends unsolicited Neighbor Advertisements in order 264 to (unreliably) propagate new information quickly." 266 Propagating information about new GUA as quickly as possible is 267 exactly what is required to solve the problem outlined in this 268 document. Therefore the host might send an unsolicited NA to 269 advertize its GUA as soon as the said address enters Optimistic or 270 Preferred state. The NA should include the target link-layer address 271 option. To ensure that all first-hop routers receive the 272 advertisement it could be sent to all-routers multicast address 273 (ff02::2). 275 As it's been mentioned, [RFC4861] explicitly states that receiving a 276 NA should not create a new NC entry. However the justification for 277 that requirement ("There is no need to create an entry if none 278 exists, since the recipient has apparently not initiated any 279 communication with the target.") clearly does not apply for the case 280 discussed. As per [RFC2119] "there may exist valid reasons in 281 particular circumstances to ignore a particular item, but the full 282 implications must be understood and carefully weighed before choosing 283 a different course.". Therefore routers creating a new NC entry upon 284 receiving an unsolicited NA would still be compliant with [RFC4861]. 286 It should be noted that some routing and switching platforms have 287 implemented such behaviour already. Administrators could enable 288 creating neighbor discovery cache entries based on unsolicited NA 289 packets sent from the previously unknown neighbors on that interface. 291 2.3.1.1. Pros 293 o Already implemented on some platforms. 295 o In accordance with [RFC4861]. 297 2.3.1.2. Cons 299 o Allows a malicious host to execute an ND cache exhaustion attack. 300 It's recommended that this functionality is configurable and 301 recommendations from [RFC6583] are taken into account. 303 o Requires hosts to send unsolicited NA (changes to the hosts). 305 o Some wireless devices are known to fiddle with ND packets and 306 perform various non-obvious forms of ND proxy actions. In some 307 cases unsolicited NAs might not even reach the routers. 309 2.3.2. Host Sending NS to the Router Address from Its GUA 311 The host could force creating a STALE entry for its GUA in the router 312 ND cache by sending the following Neighbor Solicitation message: 314 o The NS source address is the host GUA. 316 o The Source Link-Layer Address option contains the host link-layer 317 address. 319 o The target address is the host default gateway address (the 320 default router address the host received in the RA). 322 The main disadvantage of this approach is that it would not work if 323 the GUA the host needs to advertise is still in the Optimistic state. 324 The section 2.2 of [RFC4429] explicitly prohibits sending Neighbor 325 Solicitations from an Optimistic Address. 327 2.3.2.1. Pros 329 o Router implementations which follow recommendations made in 330 [RFC6583] might prioritize responding to NS packets to own 331 addresses. 333 2.3.2.2. Cons 335 o Does not work for Optimistic addresses (see section 2.2 of 336 [RFC4429]). 338 o If first-hop redundancy is deployed in the network, the NS would 339 reach the active router only, so all backup routers (or all active 340 routers ex. one) would not get their neighbor cache updated. 342 o Some wireless devices are known to fiddle with ND packets and 343 perform various non-obvious forms of ND proxy actions. In some 344 cases unsolicited NAs might not even reach the routers. 346 2.3.3. Host Sending Router Solicitation from its GUA 348 The host could send a router solicitation message to 'all routers' 349 multicast address, using its GUA as a source. If the host link-layer 350 address is included in the Source Link-Layer Address option, the 351 router would create a STALE entry for the host GUA (see the section 352 6.2.6 of [RFC4861]). However this approach can not be used if the 353 GUA is in optimistic state: the section 2.2 of [RFC4429] explicitly 354 prohibits using an Optimistic Address as the source address of a 355 Router Solicitation with a SLLAO as it might disrupt the rightful 356 owner of the address in the case of a collision. So for the 357 optimistic addresses the host can send an RS without SLLAO included. 358 In that case the router may respond with either a multicast or a 359 unicast RA (only the latter would create a cache entry). 361 2.3.3.1. Pros 363 o Unlike NS packets, RS packets would reach all routers on link, 364 allowing all routers to update their neighbor caches and 365 preventing packet loss in case of asymmetric routing. 367 2.3.3.2. Cons 369 o As for the Optimistic addresses SLLAO can not be included into RS 370 packets, the cache entry for the optimistic address would be 371 created only if the router sends solicited RAs as unicast. In 372 addition, there might be a random delay between receiving an RS 373 and sending a unicast RA back (and creating a cache entry) which 374 might undermine the idea of creating the cache entry proactively. 376 o Some wireless devices are known to fiddle with ND packets and 377 perform various non-obvious forms of ND proxy actions. In some 378 cases RSes might not even reach the routers. 380 2.4. Routers Populating Their Caches by Gleaning From Neighbor 381 Discovery Packets 383 If hosts do not send unsolicited NAs upon configuring new addresses 384 as described above the routers may be able to learn about new address 385 by gleaning from the DAD Neighbor Solicitation messages. The router 386 could listen to all solicited node multicast address groups and upon 387 receiving a Neighbor Solicitation from the unspecified address search 388 its Neighbor Cache for the solicitation's Target Address. If no 389 entry exists the router may create an entry for and set it's 390 reachability state to 'INCOMPLETE'. Then the router can start the 391 address resolution for the new entry. 393 2.4.1. Pros 395 o No changes required on hosts. 397 2.4.2. Cons 399 o Routers would receive all multicast Neighbor Discovery packets 400 which might negatively impact the routers CPU. 402 o If the router starts the address resolution as soon as it receives 403 the DAD Neighbor Solicitation the host might be still performing 404 the DAD and the target address might be tentative. In that case 405 the host SHOULD silently ignore the received Neighbor Solicitation 406 from the router as per the Section 5.4.3 of [RFC4862]. Such race 407 condition scenario would prevent the router to learn the new 408 address. 410 2.5. Initiating Hosts2Routers Communication 412 Every time the host configures a new GUA (when the address enters the 413 Optimistic state or, if the optimistic DAD is not used, as soon as it 414 changes the state from tentative to preferred) the host can a ping or 415 traceroute packet to the default gateway LLA. As the RTT to the 416 default gateway is lower than RTT to any off-link destinations it's 417 quite likely that the router would start the neighbor discovery 418 process for the host GUA before the first packet of the returning 419 traffic arrives. There are pretty good chances that the process 420 would be completed before the actual data traffic reaches the router. 422 2.5.1. Pros 424 o As data packets are involved, there is no potential impact caused 425 by smart wireless infrastructure performing ND proxy. 427 o Full compliance with existing standards. 429 2.5.2. Cons 431 o Data packets to the router LLA could be blocked by security policy 432 or control plane protection mechanism. 434 o Maximum overhead for routers control plane (in addition to 435 processing ND packets, the data packet needs to be processed as 436 well). 438 o If the first hop redundancy is implemented in the network the host 439 ping/traceroute packet would reach the active router only. All 440 backup routers would not receive it and therefore would not start 441 populating the cache. So in the case of asymmetric traffic flow 442 (packets leave the network via one router while the return flow is 443 going via another) the backup router(s) still would not have the 444 cache entry. (A hacky way to overcome this limitation would be 445 sending ping/traceroute packet to 'all routers' ff02::2 multicast 446 address). 448 2.6. Tweaking Probing Algorithms 450 While tweaking the probing logic on devices might make the problem 451 less visible it would be still desirable to avoid packet loss 452 everytime the new GUA is used by a host. It would be quite tricky to 453 adjust every probing algorithm to find the right balance between 454 prompt detection of network connectivity and false positives in 455 IPv6-only mode. 457 2.7. Routers Buffering More Packets 459 Another way to mitigate the issue, at least partially, would be 460 increasing the number of packets the router could buffer while 461 performing the neighbor discovery process for the INCOMPLETE cache 462 entry. However it would be against recommendations made in the 463 section 7.2.2 of [RFC4861] and [RFC6583]. 465 2.7.1. Pros 467 o Does not require changes on hosts. 469 2.7.2. Cons 471 o This approach makes the routers even more vulnerable to attack 472 vectors described in [RFC6583]. In particular, it would amplify 473 the impact of any scanning attack. 475 o Against the recommendations from the section 7 of [RFC6583]. 477 o Requires router vendors support. 479 3. Recommendations 481 o Hosts SHOULD send at least one unsolicited NA packet to all- 482 routers multicast address (ff02::2) as soon as one of the 483 following events happens: 485 * (if Optimistic DAD is used): a new Optimistic GUA is assigned 486 to the host interface. 488 * (if Optimistic DAD is not used): a GUA changes the state from 489 tentative to preferred. 491 o Routers SHOULD have a configuration knob to enable creating ND 492 cache entry upon receiving unsolicited NAs on a specific 493 interface. This document does not change the behavior if the ND 494 cache entry already exists when receiving an unsolicited NA. 496 As the recommendations include modification to Neighbor Discovery 497 state machine defined in [RFC4861] and hosts behaviour, they are 498 discussed in a separate Standart track document draft-linkova-6man- 499 grand. 501 4. IANA Considerations 503 This memo asks the IANA for no new parameters. 505 5. Security Considerations 507 See the Security Considerations section of draft-linkova-6man-grand. 509 6. Acknowledgements 511 Thanks to the following people (in alphabetical order) for their 512 review and feedback: Lorenzo Colitti, Igor Gashinsky, Tatuya Jinmei, 513 Erik Kline, Warren Kumari, Michael Richardson, Pascal Thubert, 514 Loganaden Velvindron, Eric Vyncke. 516 7. References 518 7.1. Normative References 520 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 521 Requirement Levels", BCP 14, RFC 2119, 522 DOI 10.17487/RFC2119, March 1997, 523 . 525 [RFC4291] Hinden, R. and S. Deering, "IP Version 6 Addressing 526 Architecture", RFC 4291, DOI 10.17487/RFC4291, February 527 2006, . 529 [RFC4429] Moore, N., "Optimistic Duplicate Address Detection (DAD) 530 for IPv6", RFC 4429, DOI 10.17487/RFC4429, April 2006, 531 . 533 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 534 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 535 DOI 10.17487/RFC4861, September 2007, 536 . 538 [RFC4862] Thomson, S., Narten, T., and T. Jinmei, "IPv6 Stateless 539 Address Autoconfiguration", RFC 4862, 540 DOI 10.17487/RFC4862, September 2007, 541 . 543 [RFC6775] Shelby, Z., Ed., Chakrabarti, S., Nordmark, E., and C. 544 Bormann, "Neighbor Discovery Optimization for IPv6 over 545 Low-Power Wireless Personal Area Networks (6LoWPANs)", 546 RFC 6775, DOI 10.17487/RFC6775, November 2012, 547 . 549 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 550 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 551 May 2017, . 553 [RFC8305] Schinazi, D. and T. Pauly, "Happy Eyeballs Version 2: 554 Better Connectivity Using Concurrency", RFC 8305, 555 DOI 10.17487/RFC8305, December 2017, 556 . 558 [RFC8505] Thubert, P., Ed., Nordmark, E., Chakrabarti, S., and C. 559 Perkins, "Registration Extensions for IPv6 over Low-Power 560 Wireless Personal Area Network (6LoWPAN) Neighbor 561 Discovery", RFC 8505, DOI 10.17487/RFC8505, November 2018, 562 . 564 7.2. Informative References 566 [RFC4941] Narten, T., Draves, R., and S. Krishnan, "Privacy 567 Extensions for Stateless Address Autoconfiguration in 568 IPv6", RFC 4941, DOI 10.17487/RFC4941, September 2007, 569 . 571 [RFC6583] Gashinsky, I., Jaeggli, J., and W. Kumari, "Operational 572 Neighbor Discovery Problems", RFC 6583, 573 DOI 10.17487/RFC6583, March 2012, 574 . 576 Author's Address 578 Jen Linkova 579 Google 580 1 Darling Island Rd 581 Pyrmont, NSW 2009 582 AU 584 Email: furry@google.com